Automatic Unsupervised Ensemble Outlier Model Selection--Extended Version

In three linesMetaEns, an automated framework for selecting unsupervised outlier detection model ensembles. Uses labeled meta-datasets to predict marginal ensemble gains and applies greedy selection with adaptive early stopping. Tested on 39 real-world datasets: outperforms baselines in average precision while using fewer models.

Summary generated by Claude — human-verified